Explain simulation architecture choices as possible worlds, not implementation toggles

A useful decision menu for an emergent simulation should state, for every option: what authoritative quantities exist, what each actor can observe, what action it can physically fund, how the world resolves that action, what behavior becomes reachable or impossible, and which numeric controls remain available to the world author. Labels such as strict, hybrid, greedy, or global are insufficient because they hide the causal consequences. A strong default is frozen observer distributions, bounded Pareto candidate generation, actor-local scalarization, one conserved allocation frame, truth-only settlement, and removable narrative readouts. This keeps outcomes emergent while still giving authors control through capacities, priors, noise, costs, response curves, and initial conditions.